Clinical Trial Perspective
CLINICAL
INVESTIGATION Statistical design and analysis issues for the ARDS Clinical Trials Network: the Coordinating Center perspective Clin. Invest. (2012) 2(3), 275–289 We describe the statistical design principals for clinical trials conducted by the Acute Respiratory Distress Syndrome (ARDS) Clinical Trials Network that is funded by the National Heart Lung and Blood Institute. These include the definition of ARDS used by the network, the choice of measures of treatment efficacy, the choice of sample size, eligibility criteria, the choice of data to collect, stratification and randomization, the use of factorial trials and coenrollment of patients on more than one trial, early stopping for futility and efficacy, the choice of control group, the control of cointerventions, treatment administration and covariate adjustment.
David A Schoenfeld*, Douglas Hayden, Cathryn Oldmixon, Nancy Ringwood & B Taylor Thompson Department of Biostatistics, 50 Staniford Street, Suite 560, Massachusetts General Hospital, Boston, MA 02114, USA *Author for correspondence: Tel.: +1 617 726 6111 E-mail:
[email protected]
Keywords: acute lung injury • coenrollment • cointerventions • early stopping rules • factorial clinical trial • hospital mortality • on target • usual care • ventilator-free days
The Acute Respiratory Distress Syndrome (ARDS) Clinical Network was established by the the National Heart Lung and Blood Institute in 1994 to develop and conduct clinical trials to prevent, treat and improve the outcome of patients with acute lung injury (ALI), ARDS and possibly other related critical illnesses. The network has refined its approach to clinical trials over the past 16 years. The purpose of this article is to describe the network’s approach to the major issues in study design and analysis, which are particular to trials of acute life-threatening diseases, such as ARDS. The National Heart, Lung, and Blood Institute created the network by requesting proposals for each site that would enter patients and for a separate ‘Clinical Coordinating Center’ that would organize the network, provide clinical coordination and handle statistical and data management issues. This article focuses on the central statistical decisions made by the network; many of which were discussed in the original Coordinating Center application. Where possible we use the data accumulated by the network to evaluate these decisions. Other aspects of the network’s organization are described elsewhere [1]. Table 1 summarizes the network studies to date [2–10]. The definition of ARDS
The original Coordinating Center proposal noted that there was not a single definition of ARDS in the literature [11–13], although all of the definitions agreed that ARDS needed to be defined as lung injury that was severe, acute and had no other obvious etiology. One of the first decisions of the network was to extend its scope from ARDS to ALI, with ALI defined as the first occurrence of a PaO2/FiO2